Bare, Litter and Climate: The Transition and Correlations

Author

A DSAN Student

Introduction

Growing up in a coastal city in China, I had never seen bare grounds, deserts, or anything other than sycamore trees, big lawns, and beaches until I visited the mid-west part of China, where the Qin Shi Huang’s Buried Terracotta Army was discovered: Xi’an. The city is famous for its historical remains, food, and of course, the desert in the middle of the city. I remember the moment I got off the plane, the hot and dry air struck me, and I was blown away by the desert and the weather, thousands of miles away from my home. I had the privilege of growing up and living by the ocean, but not all people in China, and worldwide, are so lucky that they have access to fresh water and high-vegetation-covered surroundings. 12-year-old-old me has wondered, what caused the form of desert and bare ground: is its location, or climate change?

Thanks to DSAN, I got the chance to answer (if not, at least research) the questions that young me brought up years ago.

Finding the Correlations.

The area that this story studies is in the mid-west of the US. In the lower right corner of Utah, there is a natural reserve called “Natural Bridges National Monument”, which is famous for the weird-looking stones, especially the one giant arch that looks like a huge bridge (like the key bridge that connects Georgetown and Rosslyn). Utah, which is also the top state on my travel wish list, has a dry, semi-arid, and desert climate. It is one of the driest states in the country with one of the lowest relative humidity percentages.

In the reserve, the author of the dataset categorized the vegetation coverage of the reserve into Tree Canopy, Bare Ground, Herbaceous Plants, Plant Litter, and Shrub. All of the categories make sense, except for one: plant litter. According to Google, plant litter is dead plant material that has fallen to the ground, such as leaves, twigs, flowers, fruits, and bark. It’s also known as litterfall the litter layer, or O horizon. It surprised me that the plant waste counts as vegetation coverage, since in the dry weather of Utah, I assume that the plant waste will just turn into dry leaves, then leaf dust in no time.

On second thought: what happens after the plant litter? Will there be new plants, or the area turn into bare ground? My theory is: that the plants have to die first, then their “bodies” will turn into plant waste, and the waste dissolves leaving nothing but bare ground behind.

To find evidence to support my theory, I took a look into the confusion matrix, which is the correlation matrix between each variable in the dataset. Statistically, based on the historical data (the year 1980 to 2018), the Plant Litter coverage percentage is positively correlated with Bare Ground coverage, as well as the Shrub coverage, with high correlation scores. Meanwhile, Bare and Shrub coverage has a high negative correlation.

Those statistics mean:

  1. Higher Shrub coverage → higher Litter coverage

  2. Higher Litter coverage → higher Bare coverage

  3. Higher Shrub coverage → lower Bare coverage

Code 
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt

history = pd.read_csv('../data/NABR_historic.csv')
df = pd.read_csv('../data/nearterm_data_2020-2024.csv')

columns_to_drop = ['long', 'lat', 'year', 'TimePeriod', 'RCP', 'scenario']
his_eda = history.drop(columns=columns_to_drop)
corr_matrix = his_eda.corr()

corr_matrix = history[['treecanopy', 'Ann_Herb', 'Bare', 'Herb', 'Litter', 'Shrub']].corr()
plt.figure(figsize=(10, 8))
sns.heatmap(corr_matrix, annot=True, fmt='.2f', cmap='coolwarm', square=True)
plt.title('Correlation Matrix')
plt.show()

Fig 1: Correlation Matrix

What Causes the Transition?

Now that I know what is the “afterlife” of a shrub, going back to the question that 12-year-old me would ask: What caused the transition? What aspects contribute to the transition, is it climate, rainfall, or anything else?

(You may think what kind of 12 years-old could think of the word “transition”, or “climate”: the kind of 12 years-old that will get into DSAN and work as a data scientist :) )

Code 
corr_matrix = data[['Bare', 'Litter', 'Shrub','DrySoilDays_Summer_whole', 'Evap_Summer',
       'ExtremeShortTermDryStress_Summer_whole', 'FrostDays_Winter',
       'NonDrySWA_Summer_whole', 'PPT_Winter', 'PPT_Summer', 'PPT_Annual',
       'T_Winter', 'T_Summer', 'T_Annual', 'Tmax_Summer', 'Tmin_Winter',
       'VWC_Winter_whole', 'VWC_Spring_whole', 'VWC_Summer_whole',
       'VWC_Fall_whole']].corr()
plt.figure(figsize=(10, 8))
sns.heatmap(corr_matrix, annot=True, fmt='.2f', cmap='coolwarm', square=True)
plt.title('Correlation Matrix')
plt.show()

Fig 6: Correlation Matrix between Bare, Shrub, Litter, and climate variables.

Repeating the previous research process and starting by examining the statistical relations: it seems like, comparably, those variables: ‘NonDrySWA_Summer_whole’, ‘VWC_Winter_whole’, ‘VWC_Spring_whole’, ‘VWC_Summer_whole’, ‘VWC_Fall_whole’, are positively correlated with Litter coverage, and negatively correlated with Bare Ground coverage.

NonDrySWA_Summer_whole: Represents the non-dry soil water availibility during summer; mean soil water availibility during days with the soil >-3.9MPa during summer in total centimeters

VWC_Season_whole: Represents the volumetric water content in the top 50cm of soil during winter/summer/spring/fall; median volumetric water content in the top 50cm of soil during winter in m^3 / m^3.

This finding indicates that the transition between shrub → litter → bare ground is correlated with soil water availability in the summer and year-round median volumetric water content in the top surface of the soil.

Code 
import plotly.graph_objects as go

columns_to_check = ['NonDrySWA_Summer_whole', 'VWC_Winter_whole', 'VWC_Spring_whole', 'VWC_Summer_whole', 'VWC_Fall_whole']
df_1 = data.dropna(subset=columns_to_check)
df_1 = df_1.dropna(subset=columns_to_check)
df_1 = df_1.drop_duplicates(subset=columns_to_check)
df_1 = df_1.loc[
    ((df_1['RCP'] == '4.5') & (df_1['scenario'] == 'sc22')) |
    ((df_1['RCP'] == 'historical') & (df_1['scenario'] == 'sc1'))
]

mean_data_by_year = df_1.groupby('year')[columns_to_check].mean().reset_index()

df_1.to_csv('../data/df_1.csv', index=False)

columns_to_check_2 = ['Bare', 'Shrub']
mean_data_by_year_2 = data.groupby('year')[columns_to_check_2].mean().reset_index()


fig = go.Figure()

fig.add_trace(go.Scatter(x=mean_data_by_year_2['year'], y=mean_data_by_year_2['Bare'], 
                         mode='lines', name='Bare', 
                         line=dict(color=color_mapping['Bare'])))

fig.add_trace(go.Scatter(x=mean_data_by_year_2['year'], y=mean_data_by_year_2['Shrub'], 
                         mode='lines', name='Shrub', 
                         line=dict(color=color_mapping['Shrub'])))

fig.add_trace(go.Scatter(x=mean_data_by_year['year'], y=mean_data_by_year['NonDrySWA_Summer_whole'], 
                         mode='lines', name='NonDrySWA_Summer_whole', 
                         line=dict(color=color_mapping['NonDrySWA_Summer_whole']), visible='legendonly', yaxis = 'y2'))

# Add VWC_Winter_whole to the secondary y-axis
fig.add_trace(go.Scatter(x=mean_data_by_year['year'], y=mean_data_by_year['VWC_Winter_whole'], 
                         mode='lines', name='VWC_Winter_whole', 
                         line=dict(color=color_mapping['VWC_Winter_whole']), visible='legendonly', yaxis = 'y2'))

# Add VWC_Spring_whole to the secondary y-axis
fig.add_trace(go.Scatter(x=mean_data_by_year['year'], y=mean_data_by_year['VWC_Spring_whole'], 
                         mode='lines', name='VWC_Spring_whole', 
                         line=dict(color=color_mapping['VWC_Spring_whole']), visible='legendonly', yaxis = 'y2'))

# Add VWC_Summer_whole to the secondary y-axis
fig.add_trace(go.Scatter(x=mean_data_by_year['year'], y=mean_data_by_year['VWC_Summer_whole'], 
                         mode='lines', name='VWC_Summer_whole', 
                         line=dict(color=color_mapping['VWC_Summer_whole']), visible='legendonly', yaxis = 'y2'))

# Add VWC_Fall_whole to the secondary y-axis
fig.add_trace(go.Scatter(x=mean_data_by_year['year'], y=mean_data_by_year['VWC_Fall_whole'], 
                         mode='lines', name='VWC_Fall_whole', 
                         line=dict(color=color_mapping['VWC_Fall_whole']), visible='legendonly', yaxis = 'y2'))

# Update the layout to include secondary y-axis
fig.update_layout(
    title='Volumetric Water Content Trends Over Time',
    xaxis=dict(title='Year', showgrid=False),
    yaxis=dict(title='m^3', showgrid=False),
    yaxis2=dict(title='cm', showgrid=False, overlaying='y', side='right'),
    legend=dict(x=1.10, y=1, xanchor='left', yanchor='top', bgcolor='rgba(255, 255, 255, 0.5)', bordercolor='black', borderwidth=1),
    hovermode='x unified',
    plot_bgcolor='rgba(200, 182, 193, 0.3)'
)


# Show the figure
fig.show()

Fig 7: Volumetric Water Content Trends Over Time

To further prove the hypothesis, I noticed that the Bare ground coverage trend is more closely correlated with “non-dry soil water availability during summer” (NonDrySWA_Summer_whole), and “volumetric water content in the top 50cm of soil during summer” (VWC_Summer_whole), compares to other seasons.

So, my initial conclusion is transition between shrub → litter → bare ground is correlated with soil water availability and median volumetric water content in the top surface of the soil in the summertime.

Futher Investigation: Location’s Impact.

Although water availability is crucial for the shrub’s transition, other variables may have an impact on the process. The location’s impact could be significant and inevitably moving from east to west across the country, it could also mean something even within one natural reserve.

Since we already discovered the correlations between the shrub’s transition and water availability, the next step is to take a look at the location’s impact on the vegetation coverage.

Code 
import pandas as pd
import folium
from folium.plugins import HeatMap
import ipywidgets as widgets
from ipywidgets import interact
import branca.colormap as cm
from folium import IFrame, Popup

# Calculate the central coordinates for the initial map view
center_lat = data['lat'].mean()
center_long = data['long'].mean()

def create_heatmap(veg_type, color):
    # Filter the data
    heat_data = [[row['lat'], row['long'], row[veg_type]] for index, row in data.iterrows() if not pd.isnull(row[veg_type])]

    # Normalize the data to the range 0-1 for the heatmap
    heat_data_normalized = [[d[0], d[1], d[2] / 100] for d in heat_data]

    # Define the color map for the legend
    colormap = cm.LinearColormap(colors=['white', color], vmin=0, vmax=100, caption=f'{veg_type} Coverage (%)')

    # Create a map centered around the calculated mean latitude and longitude
    m = folium.Map(location=[center_lat, center_long], zoom_start=13)  # Adjust zoom_start as needed

    # Add the heatmap layer with a custom gradient
    HeatMap(heat_data_normalized, gradient={0.0: 'white', 1.0: color}, min_opacity=0, max_opacity=1).add_to(m)

    # Add the colormap (legend) to the map
    colormap.add_to(m)

    title_html = f'''
        <div style="position: fixed; 
                    bottom: 10px; left: 10px; width: 300px; height: 60px; 
                    background-color: white; opacity: 0.7; z-index:9999; font-size:16px;
                    border:2px solid grey; padding: 10px;">
            <b>{veg_type} Coverage Heatmap</b><br>
        </div>
    '''
    
    # Add the title and subtitle to the map
    m.get_root().html.add_child(folium.Element(title_html))

    # Display the map
    return m

interact(create_heatmap, veg_type='Bare', color=color_mapping['Bare'])

Fig 8: Bare Coverage Heatmap

<function __main__.create_heatmap(veg_type, color)>
Code 
interact(create_heatmap, veg_type='Shrub', color=color_mapping['Shrub'])

Fig 9: Shrub Coverage Heatmap

<function __main__.create_heatmap(veg_type, color)>
Code 
interact(create_heatmap, veg_type='Litter', color=color_mapping['Litter'])

Fig 10: Litter Coverage Heatmap

<function __main__.create_heatmap(veg_type, color)>

Surprisingly, all the three vegetation: Bare, Shrub, and Litter, are denser on the northwest side of the reserve.

Some ideas for further investigations are:

Elevation differences: Is the elevation descending from northwest to southeast? This may cause the water availability in the top surface of the soil to be lower for the northwest than the southeast. Since, if the elevation is higher, this means higher floor temperature due to the direct sunlight.

Wind: Is the wind coming from the northwest to southeast? If so, the wind could take the moisture, also prevent taller trees (taller than shrubs) from growing, since taller trees need more water from the soil and be able to stand up to the wind.

Human/Animal Activities: Is the allocation of vegetation coverage manipulated by humans? Or are they the habitat of some animals in the reserve? If so, I will need more data and information to prove my assumption.

Conclusion

As I was saying before, I have the privilege to live on the coastline, and never have to worry about water, having fresh groceries, and seeing greens everywhere I go. That’s why I believe everyone (who is financially able to, of course) must travel, and explore the world: not only traveling to scenery places, but also places like Xi’an or Utah, where the culture is rich and stones are magnificent, but the climate is not very desirable. Only this way one could appreciate the accessibility to fresh water, and raise awareness of issues like water saving, desertation, climate change, etc.

I still remember a story I read in middle school, about a project in China that reversed desertation, planting trees in the desert by first changing the sand into something that could contain water. It took the residents over 20 years to accomplish the task and planted thousands of trees on the area, where it was originally thousands of anchors of deserts. During the process, there were failures and desperation, but in the end, the outcome was breathtaking. Not until then I realized, how serious dissertation is in China, and worldwide, and it occurred to me that I am so lucky to be surrounded by all the sycamore trees. Now, looking at the dataset, on top of finding correlations between variables and coming to a conclusion, I feel more related to the fact that the majority of the vegetation coverage is Bare ground (about 55% each year) and wish I could contribute something to make the earth greener again.

References

Dataset: https://www.sciencebase.gov/catalog/item/61a6952fd34eb622f6978d9f

Picture: National Park Service